it depends on your prompting. LLMs can generate beautiful code, or absolute slop.